In this study, multitemporal landslide layers combined with landscape metrics, spatial analysis, and logistic regression were used to monitor spatiotemporal changes in forest restoration and identify key factors promoting natural restoration. The total area of landslides caused by Typhoon Morakot markedly decreased within the first 2 years,
followed by a gradual decline. Ecological regeneration peaked in 2012 and then stabilized. Logistic regression indicated landscape indicators such as CONTIG (a patch metric), distance from the landslide boundary, AREA (a patch metric), elevation, and slope gradient as key factors influencing natural restoration. These findings highlight the importance of seed source availability in the natural restoration of landslide areas. Landslide areas with more dispersed landscapes,larger areas, lower elevations, and gentler slopes appear to be more conducive to natural restoration. The results may guide future forest management.
Key Words: landslide, natural restoration, spatial autocorrelation, landscape metric, logistic regression